Acoustic micromachining of three-dimensional surfaces for biological applications

Abstract

We present the use of an accessible micromachining technique (acoustic micromachining) for manufacturing micron-feature surfaces with non-discretely varying depth. Acoustic micromachining allows for non-photolithographic production of metal templates with programmable spatial patterns and involves the use of standard acoustic, cutting and electroplating equipment for mass production of vinyl records. Simple 3D patterns were transferred from an acoustic signal into working nickel templates, from which elastic polymer molds were obtained, featuring deep surface grooves and non-discrete (smooth) variations in the z-dimension. Versatility and applicability of the method is demonstrated in obtaining microfluidics structures, manufacturing high-surface area wavy polymer fibers, assembly of cell networks on scaffolds with 3D topography, and microcontact printing of proteins and cells.

---

Introduction

Microfabrication and micropatterning have found wide application in elucidating basic cellular responses in biological systems.^1,2 The ability to engineer micron-scale structures with multiple heights, i.e. three-dimensional (3D) structures, permits one to study 3D interactions between cells and their extracellular environment, not otherwise attainable in traditional 2D flat culture systems. Stringent 3D control is even more important for microelectromechanical (MEMS) and microfluidic applications where precise geometry is critical. Traditional photolithography produces features of uniform height by virtue of the photomask used in the creation of these structures. Multiple masks can be used to create multi-level surfaces with certain restrictions, and recently several modified photolithographic techniques have been published providing fine discrete steps in feature height.^3,4 To circumvent the problem of clean rooms and toxic photoresist materials required for photolithography, there have been attempts to create templates by casting self-assembled structures or optical diffraction gratings⁵ but such solutions are limited to simple repeated patterns of small height. Truly programmable 3D surfaces can be obtained using computer-numeric-controlled (CNC) milling techniques, selective laser ablation (SLA), or selective laser sintering (SLS). Although CNC milling is a more flexible method than the technique we describe below, it suffers from its discrete, coarse steps. SLA and SLS both are prohibitively expensive, have a limited repertoire of materials that are compatible with the process, and require complex programming.

All of the techniques discussed so far produce discrete (in the z-dimension) surface features. In this paper, we explore the capabilities of an established method and technology—the vinyl record manufacturing industry—to produce custom-designed inexpensive templates suitable for controlled cell growth and general biological applications. The Columbia Long-Playing (LP) microgroove recording technique was established in 1948, and soon after that became the major music recording technology until the introduction of newer media.^6,7 The LP production process offers micron-scale capabilities, deep feature sizes, and a non-discrete (smooth) variation in depth and width in response to the audio signal, effectively producing analog 3D patterns. We demonstrate the use of the method in obtaining microfluidics structures, manufacturing high-surface area wavy polymer fibers, for assembly of cell networks on scaffolds with 3D topography, and microcontact printing of proteins and cells. We also discuss the advantages and limitations associated with the proposed technique.

Experimental

Acoustic micromachining

Topographic encoding of the acoustic signal. An audio signal is transferred into a topographic image by cutting V-grooves at constant angular velocity onto a soft lacquer layer. The movement of the cutting head (typically a sapphire with a 90° finish) is modulated electromagnetically and independently by the signal through a left and a right amplifier, as outlined schematically in Fig. 1. Alternative shapes of the stylus, including sharper angles are available but are considered exotic in the standard production process. For technical and historical reasons, the movement encoded by both channels is along the normal to a side wall and is inverted between the two channels. Therefore, identical (in-phase) signals for the two amplifiers (essentially a monaural record) result in a lateral motion, while pure vertical motion is achieved by inverting one channel, or in the case of regular periodic signals shifting one channel 180° anti-phase to the original signal. The depth of the grooves is adjustable and determines the dynamic range, e.g. volume, of the signal; the distance between the grooves is determined by the degree of packing (a tradeoff of dynamic range and capacity of the record), and can be optimized during the recording process using a groove computer or set manually at the start of the cutting procedure. The cutting speed has to be properly adjusted and stylus heated to achieve clean cuts and avoid the formation of “burrs” of material.

Fig. 1 The acoustic encoding process. A. The movement of the cutting head (stylus) along the normals to the walls of the groove is modulated by the encoded signal for the two amplifier channels, Ch 1and Ch 2. B. For a monaural record, both channels are driven by in-phase signals, but due to a (historically-introduced) inversion of one of them, a net lateral motion is produced (see resultant vector). C. For a stereo record, both channels are driven by anti-phase signals (shifted by 180°), thus the stylus is moving in the vertical (z-direction).

Production of metal templates. For mass production purposes, the lacquer is used as a template and destroyed in the process of creating a nickel-alloy template having a negative image of the lacquer which itself can be used in producing a metal positive impression. Silver nitrate is sprayed onto the surface to create a conductive layer. Then a galvanic electroforming process is applied with the silver layer acting as a cathode, and nickel from the bath being deposited onto the silver layer. The nickel layer, a negative of the lacquer, is removed and becomes a Master. From this Master, further copies are produced: a Mother (a positive replica of the lacquer) and a Stamper (a negative of the lacquer), which is used, as the name implies, to stamp out the actual vinyl LPs.

Custom design and limitations of the process. The metal Mother and Stamper are durable and ideal for further molding of various materials. For a conventional cutting head, the obtained features are trapezoidal grooves with triangular peaks (V-peaks) (from the Mother) and V-grooves (from the analogous Stamper). The depth can be adjusted from approximately 6 µm (0.25 mils‡) up to the thickness of the lacquer (about 150 µm). The groove-to-groove spacing is adjustable, with a lower limit about 30 µm. Parallel V-grooves with a constant depth can be obtained by recording ‘silence’, i.e. no frequency modulation; changing the volume/depth while recording will achieve spatial depth variations.

The frequency response for this process is in the range of 7 Hz to 25 kHz with a maximum sustained frequency of about 16 kHz. During production, a constant angular velocity (variable linear velocity) is maintained (33.3 or 45 rpm for the current standards), thus feature size decreases as a function of the radial position during the cutting process. Given the internal (378 mm) and external (900 mm) circumference of a 12″ record, the range of linear velocities obtained for a 33.3 rpm LP is from 210 to 500 mm s⁻¹. Fig. 2A outlines the spatial features obtainable as a function of the modulating frequency. Periodic structures below 100 µm can be obtained for frequencies >2 kHz; the smallest periodic structures are in the range of 10 µm for the maximum attainable modulation frequency. Fig. 2B shows a computer simulated V-groove encoded using a 500 Hz square wave monoaural signal (after low-pass filtering with a Butterworth filter, cut-off frequency15 kHz). The guiding equations are given in the Appendix; the Matlab program for groove cutting simulations and visualization is available as electronic supplementary information.†

Fig. 2 Feature size and patterns limitations. A. Size (wavelength) of periodic features that can be produced if sinusoidal signals of prescribed frequencies (x-axis). The three curves refer to three radial positions on a standard 12″ disk—outside radius, mid-radius and inner radius—which result in different wavelengths because of the variable linear (constant angular) stylus speed during manufacture. B. Computer-simulated cut-out groove pattern by a 500 Hz square wave after low-pass filtering to account for the stylus size. Inset zooms in on the edge of the pattern; shown is also the original versus the filtered version of the signal.

Polymer molding and surface characterization

Polydimethylsiloxane or PDMS (Sylgard 184 from Dow Corning, Midland, MI) was mixed in the usual ratio 1 : 10, molded out of the metal templates to introduce 3D topography, and baked for 2 h at 60 °C.⁸ Mother templates yielded V-grooves in the PDMS while Stamper templates formed V-peaks. The obtained patterns (cut in 1 cm² pieces) were used in microfluidic applications, to grow cells after protein coating, or as protein stamping surfaces. Alternatively, a mixture of cellulose acetate and acetone (0.05 g ml⁻¹, both from Aldrich, Milwaukee, WI) was poured over the metal templates in a pre-determined amount, which after acetone evaporation resulted in the formation of discrete separable fibers.⁹ In some cases, Congo red (Sigma, St. Louis, MI) dye was used to better visualize molded structures or fluid patterns. Surfaces were imaged directly or in a cross-section using optical microscopy (Nikon TS100) or scanning electron microscopy (SEM).

Cell culture, cell assembly, and cell imaging

Primary cardiomyocyte culture was prepared as described previously.^8,10 Briefly, the hearts of 3 days old Sprague-Dawley rats were digested enzymatically using trypsin and collagenase at 1 mg ml⁻¹ (Worthington Biochemical, Lakewood, NJ). Cardiomyocytes (CM) were collected after centrifugation and preplating, and seeded on fibronectin-coated (50 µg ml⁻¹, BD Biosciences, Franklin Lakes, NJ) PDMS surfaces. In this process, cardiac fibroblasts (CF) were separated during the preplating step and cultured on fibronectin-coated surfaces. Cells were kept in a CO₂ incubator in Medium 199 (Invitrogen, Carlsbad, CA) at 37 °C.

For cell patterning experiments, 3D PDMS stamps were used for microcontact printing of fibronectin on polystyrene surfaces, and cells were plated after surface treatment with 0.5% w/v Pluronic F-127 (Sigma) to avoid non-specific attachment.¹¹

For structural characterization, cells were fixed in 3.7% formaldehyde and permeabilized with 0.02% Triton-X 100 (Sigma). Cell cytoskeleton was fluorescently stained with phalloidin–Alexa Fluor 488 (Molecular Probes, Eugene, OR) for F-actin. Nuclei of fixed cells were labeled with TOTO-3 (Molecular Probes, Eugene, OR). Structural fluorescence imaging was performed using an inverted fluorescence microscope with a 40× objective (NA 0.95) or, in some cases, using a confocal laser-scanning microscope BioRad Radiance 2000 with a 60× oil-immersion objective (NA 1.4).

Dynamic fluorescence imaging in live cardiomyocyte networks was performed to track propagation of electrical waves at >60000 locations using a voltage-sensitive dye di-8-ANEPPS (Molecular Probes), an intensified CCD camera (DAGE MTI, Michigan City, IN) and a custom-built optical setup, as described previously.⁸ Propagation maps were constructed by plotting contours of activation times over the imaged space, where closeness of contours is inversely related to the velocity of propagation.

Results and discussion

Manufacture of 3D structures by acoustic micromachining

Audio signals representing silence, 5 kHz and 10 kHz monaural or stereo modulation were used for this proof-of-principle study. The signals were encoded (cut) into a lacquer template using variable volume, effectively resulting in surfaces between 6 and 50 µm deep. The packing/spacing between grooves was varied from 60 to 500 µm, with most of the structures being 120 µm apart. The metal (Ni) surfaces of the templates were durable, easy to clean and did not introduce any adverse effects through the polymer molds on the proteins and cells used in this study. Fig. 3 shows SEM images of PDMS surfaces obtained by molding. Encoded silence (no frequency modulation) results in nearly straight grooves; in-phase (monoaural) signals result in a lateral stylus modulation and wavy (in x–y) grooves; while anti-phase (stereo) signals to the two amplifiers result in a vertical stylus motion and smooth height modulation. For the analyzed samples, the amplitude of sinusoidal undulations was within 5 µm. The radius of curvature introduced by the sinusoidal stylus motion varied between 8.6 µm and 50.3 µm, determined by the frequency of the signal and the radial position on the template. Such geometrical modulations of the surface (depth or width) can invoke a response within a single cardiomyocyte or cardiac fibroblast having typical length between 60 and 120 µm, and width between 20 and 60 µm. Table 1 summarizes parameters related to the standard record cutting process (data from personal communication). It has to be noted that the nature of the cutting process limits the complexity of the patterns that can be obtained (no small closed-loop or sharp-edge structures can be fabricated). This is in contrast to photolithography where 2D spatial patterns of extremely high complexity can be produced.

Fig. 3 Manufactured 3D surfaces imaged by scanning electron microscopy (SEM). A. Encoded silence (no frequency modulation) results in straight grooves. Shown is a polymer (PDMS) mold of V-grooves. B. In-phase (monoaural) signals to the two amplifiers result in lateral stylus modulation and wavy (in x–y) grooves. Shown are a V-grooved (left) and V-peaked (right) replica of the surfaces produced by 10 kHz sinusoidal signals. C. Anti-phase (stereo) signals to the two amplifiers result in vertical stylus motion and height modulation. Shown are V-peak (left) and V-grooved (right) replica of the surfaces produced by 5 kHz and 10 kHz sinusoidal signals. Scale bar is 100 µm.

Table 1 Summary of record cutting limitations using standard equipment

Parameter	Min.	Typ.	Max.
Unmodulated depth, z0/µm	0	35	150
Unmodulated width, w = 2z0/µm	0	70	300
Wall angle/degrees	—	90	—
Lateral deviation/µm	−½δR	—	½δR
Vertical deviation/µm	−z0	—	150 − z0
Inter-groove distance, δR/µm	w	150	—
Instantaneous freq. response	7 Hz	—	25 kHz

---

Fig. 4 demonstrates several possible applications of the micromachined structures. To obtain confined microfluidic channels, we sealed (by heating) a PDMS of a stereo signal with V-grooves to a clean coverglass surface. Congo red dye was used to color the fluid; a droplet was deposited asymmetrically on one side (upper left corner in this case, Fig. 4A) and the capillary action-induced flow in the variable height grooves was imaged. The smooth (non-discrete) variations in the diameter of the channels can be used to mimic biologically relevant problems in vitro, such as aneurisms, plaque formation, associated local turbulence, to mention a few. Microfluidics can also be used for 2D molecule and cell patterning,³ as illustrated in Fig. 4B, using a red-colored fluid pattern. Fig. 4C shows the production of wavy cellulose acetate fibers with high surface-to-volume ratio, which can be then post-processed into biodegradable structures for cell and tissue growth.⁹

Fig. 4 Applications of manufactured surfaces. A. Microfluidic application is shown, where a PDMS V-peaked mold of a 5 kHz stereo signal was attached to a glass surface and red-colored fluid was fed from the left upper corner, moving by capillary action in the grooves. B. Microfluidic surface patterning (red traces) using PDMS V-peaked replica of a monoaural 5 kHz signal. C. Polymer wavy fibers (cellulose acetate with mixed red dye) having high surface-to-volume ratio were produced by molding out of a nickel V-grooved template (10 kHz stereo). Scale bar is 50 µm.

Cell assembly and control by manufactured surfaces

The utility of the micromachining technique for cell growth was explored by using two primary cell types: cardiac fibroblasts and cardiomyocytes onto fibronectin-coated PDMS surfaces. Fig. 5A and B illustrate some results obtained with CF on 3D surfaces. As shown in previous studies,^12,13 cell guidance and alignment was easily achieved by both V-groove and V-peak scaffolds. Interestingly, CF were more likely to stretch and form transverse bridges between the peaks (Fig. 5B) despite the greater distance relative to crossing the V-grooves. Two simple examples are shown how PDMS stamps prepared from the metal templates can be used for microcontact printing of extracellular matrix proteins (fibronectin, in this case) and patterning of fibroblasts and myocytes (Fig. 5C and D).

Fig. 5 Cell growth and patterning. A and B. Fibroblasts grown on V-grooved and on V-peaked PDMS surfaces, respectively. Arrows indicate where cells bridged between the peaks; dotted lines indicate the bottom of the grooves or the top of the peaks, respectively. C. Patterning of fibroblasts after microcontact printing of fibronectin using V-grooves. D. Patterning cardiomyocytes in thin lines with V-peaks. E. Oriented (anisotropic) growth of cardiomyocytes on V-peaks. F. Fluorescently measured anisotropic propagation of electrical waves in the cardiomyocyte network, triggered from a point (*). Color represents time of activation (in ms), blue being the earliest. F-actin cytoskeleton was labeled in green in A, B and E. Scale bar is 100 µm.

Cardiomyocytes grown directly onto V-peak scaffolds maintained a characteristic angle with respect to the direction of the peaks,⁸Fig. 5E. Functional connectivity between these cells and anisotropy in the wave propagation was confirmed using fast fluorescence imaging at multiple locations in live cardiomyocyte networks (Fig. 5F). A point stimulation elicits an elliptical rather than circular pattern of propagation, dictated by the assembly of the cells. This is a characteristic natural behavior for the heart tissue.

Biological applications of acoustic micromachining

The applications outlined in Fig. 4 and 5 present only several of the numerous possibilities for these 3D surfaces. Biological applications calling for controlled cellular environments or molecule manipulation platforms include: structure–function cellular studies by the use of patterning; integration of cells with microelectromechanical devices for BioMEMS design; microfluidics solutions for molecule and cell manipulation and patterning. The need for relatively simple 3D (vs. highly complex 2D) structures is apparent in several specific biological applications: (1) enforcement of prescribed cell morphology and assembly to mimic anisotropic tissue properties, by micropatterning or “topographic guidance”; (2) fluid flow studies and effects of flow on cells, where the structure (diameter/depth) of the channels is spatially variable; (3) studies of the effects of curvature and/or spatial frequencies on cell organization, effects of 3D microenvironment on cell attachment and function, where deep surface features are needed; and (4) compartmentalization platforms for cells and molecules to study local autocrine/paracrine effects, effects of pharmacological agents etc.

Summary

Soft lithography,¹⁴ combining photolithography with polymer molding, has facilitated in an unprecedented manner the development of biological applications requiring spatial control of molecules and cells. The advantages of the technique lie with its simplicity and accessibility for the regular researcher. In the same venue, we report in this paper the utility of an even simpler non-photolithographic method, acoustic micromachining, for microscale design of surfaces and 3D structures. A simple encoding of desired topographical features into an audio signal is used, and a template is generated via an established production process and equipment. We showed how acoustic micromachining can produce features ranging from several microns to hundreds of microns suitable for various biological applications.

The major limitation of acoustic micromachining is decreased flexibility in microstructure design. Groove geometry is determined by the shape of the cutting head, fixing the width–depth correlation. The rotating platform imposes yet another restriction on possible topologies in acoustic micro-machining as most cutting lathes are designed to spiral towards or away from the center of the disk at a fixed rate precluding the design of arbitrary networks of grooves. However, many microstructure applications require relatively simple features, where acoustic micromachining can successfully complement the more sophisticated photolithography (for 2D) and other general 3D methods. Geometry-encoding “music” can help effectively address current biological questions in a simple and inexpensive way.

Appendix

The equations for tracking the motion of the stylus in Cartesian coordinates (x, y, z), where z is depth, are given below for arbitrary signals supplied to the two amplifier channels, s_Land s_R.

y(t) = sin(2πνt)(R_max − νtδR)

The following symbols are used here: ν [rad s⁻¹] is the angular velocity of the stylus; δR [µm] is the inter-groove spacing—fixed or variable; R_max [µm] is the initial radial distance from the origin; and z₀ [µm] is the initial groove depth.

The electronic supplementary information† provides the Matlab code for simulations and visualization of microstructures by acoustic micromachining.

Acknowledgements

We acknowledge support from the Whitaker Foundation (RG-02-0654) to EE. We thank Lihong Yin and Chiung-yin Chung for help with cell culture, Custom Mastering (Nashville, TN) for template production and Elko Tchernev for helpful discussions.

References

1. A. Folch and M. Toner, Annu. Rev. Biomed. Eng., 2000, 2, 227.
2. G. M. Walker, H. C. Zeringue and D. J. Beebe, Lab Chip, 2004, 4, 91.
3. J. Tien, C. M. Nelson and C. S. Chen, Proc. Natl. Acad. Sci. USA, 2002, 99, 1758.
4. C. Chen, D. Hirdes and A. Folch, Proc. Natl. Acad. Sci. USA, 2003, 100, 1499.
5. Y. N. Xia, J. Tien, D. Qin and G. M. Whitesides, Langmuir, 1996, 12, 4033.
6. P. C. Goldmark, R. Snepvangers and W. S. Bachman, Proc. Inst. Radio Eng., 1949, 37, 923.
7. D. W. Gravereaux, J. Acoust. Soc. Am., 1985, 77, 1327.
8. H. Bien, L. Yin and E. Entcheva, IEEE Eng. Med. Biol. Mag., 2003, 22, 108.
9. E. Entcheva, H. Bien, L. Yin, C. Y. Chung, M. Farrell and Y. Kostov, Biomaterials, 2004, 25, 5753.
10. E. Entcheva and H. Bien, J. Biomed. Microdev., 2003, 5, 163.
11. E. Detrait, J. B. Lhoest, P. Bertrand and P. van den Bosch de Aguilar, J. Biomed. Mater. Res., 1999, 45, 404.
12. A. S. G. Curtis and C. D. Wilkinson, J. Biomater. Sci.—Polym. Ed., 1998, 9, 1313.
13. P. Clark, P. Connolly, A. S. Curtis, J. A. Dow and C. D. Wilkinson, Development, 1987, 99, 439.
14. G. M. Whitesides, E. Ostuni, S. Takayama, X. Jiang and D. E. Ingber, Annu. Rev. Biomed. Eng., 2001, 3, 335.

---

Footnotes

† Electronic supplementary information (ESI) available: Matlab code for simulations and visualization of microstructures by acoustic micromachining. See http://www.rsc.org/suppdata/lc/b4/b409478f/

‡ 1 mil = 1/1000 of an inch.

---